Summary: On January 28, 2026, the OECD published AI adoption statistics showing 36.8% of individuals used generative AI at least once in 2025. Firm adoption rose from 8.7% in 2023 to 20.2% in 2025, leaving organizations behind individual users and creating demand for packaged adoption tools. The remaining gaps are steepest by age, firm size, and sector, so interface, trust, and workflow fit matter as much as model capability.
Word came from the OECD on January 28, 2026: a fresh round of AI adoption statistics, every figure of it sourced from the ICT Access and Usage Database. Two developments deserve a mention. More than one person in three used generative AI tools during 2025, with ITIF citing a precise 36.8 percent. Firm adoption, meanwhile, doubled inside two years — 8.7% in 2023, 20.2% in 2025. No other cross-national survey spans as many countries or holds to a more consistent measurement basis, which leaves product and policy teams with rare grounding: who has already adopted these tools, and who has yet to start — arriving while most AI discussion still orbits what models can do.
Individuals: The One-Third Line Has Been Crossed
- During 2025, 36.8% of individuals used generative AI tools of the ChatGPT and Claude kind at least once.
- The Nordics are out front: Denmark, Finland, and Sweden each clear 35%.
- Cut by status, the groups sit far apart: three-quarters of students aged 16 and over, 41.1% of the employed, 36.7% of the unemployed, and only 12.5% of retirees and others outside the labor force.
Crossing one-third matters for reasons beyond the number: generative AI has stepped out of the early-adopter circle and into everyday software. That three-quarters rate among students amounts to an answer written in advance — the cohorts entering work over the coming five years will already list these tools among their basic skills. And at 12.5%, retirees and those outside the labor force mark the ground diffusion has yet to cover away from work and school.
Where the Divide Opens: Age, Education, Income
- Age is the steepest: 53.6 percentage points stand between the heaviest users and those who barely touch the tools.
- Cut by education or by income, each split produces about 21 percentage points of distance.
- Gender shows just 4.2 percentage points, plainly smaller than any other axis.
Read as one set, the figures point the same way: young generations are leading this spread, and its unevenness exceeds the PC and smartphone waves. Judged next to those earlier waves, 4.2 points of gender distance looks small; this time the break divides generations. The note for product teams is concrete — resistance sits with older people and those outside employment, and what blocks them is interface and trust, not a lack of ability.
Firms: Doubling Done, Growth Narrowing
Line the firm figures up year by year: 8.7% in 2023, 14.2% in 2024, and 20.2% by 2025. As a gain, 2025 delivers 42.4%, short of 2024’s 59.6%. Momentum has not died; a larger base naturally lowers the rate, which is the shape an S-curve takes through its middle stretch. Slower or not, the pace still added six percentage points of firms during 2025 alone, on top of the 14.2% the year inherited.
Put the two tracks together and the contrast is plain: individuals cleared one-third, firms only just reached one-fifth. Organizations trail their own staff, and the distance between the curves is the market where internal tooling and adoption services live.
Scale and Sector: How Wide the Gap, Who Leads
- Size alone accounts for a gap of nearly three times: 52.0% of large firms have adopted AI, against 17.4% of small ones.
- ICT holds first place at 57.3%, with professional, scientific, and technical services behind it at 36.8%; break ICT down further and Sweden at 87.9%, Austria at 79.9%, and Finland at 79.8% all hug the saturation line.
- The fastest movers are traditional trades off the lowest base: accommodation and food services at +62.5%, construction at +59.1%.
Look at where that growth actually lands: hospitality and construction, frontline trades with thin margins and almost no in-house ML talent. What they take up is customer communication, scheduling, and estimating rather than code, and spread of that kind comes from packaged software, not data teams.
What Product Teams Should Take Away
First, the consumer question has changed: nearly everyone who might use generative AI already has, at least once, so the contest is retention and depth rather than teaching people from zero. Second, the enterprise gap stays wide: overall adoption is only 20.2%, most companies have not made a start, and what moves them is easy onboarding and a fit with existing workflows, more than model power on its own. Third, a 53.6-point generational gap is itself a design condition — a product built for digital natives alone gives up close to half its potential market. The dataset also supports what we argued in our 2026 opening outlook: what matters this year is how deeply AI is used, not whose model makes the loudest news.